Files
song-prompt-webapp/application/background_tasks.py

84 lines
2.5 KiB
Python

from celery import shared_task
from redbeat import RedBeatSchedulerEntry
from celery import current_app as celery_app
from sqlalchemy.sql.expression import func, select, insert
from sqlalchemy.ext.automap import automap_base
from sqlalchemy import create_engine
from sqlalchemy.orm import Session
from gpt4all import GPT4All
engine = create_engine("sqlite:///words_prompts.db", pool_pre_ping=True)
Base = automap_base()
Base.prepare(engine, reflect=True)
Words = Base.classes.words
Themes = Base.classes.themes
MODEL = GPT4All(
model_name="gpt4all-falcon-q4_0.gguf",
# model_path=(Path.home() / ".cache" / "gpt4all"),
allow_download=False,
)
@shared_task
def grab_word():
while True:
with Session(engine) as word_session:
random_word = word_session.query(Words.words)
random_word = random_word.order_by(func.random()).first()
random_word = str(random_word)[4:-4]
# SYSTEM_TEMPLATE = "A single sentence based on a word."
# PROMPT_TEMPLATE = "### Instruction: {0} \n### Response: "
response = MODEL.generate(
f"Give me a writing prompt about {random_word}.",
temp=0.7,
callback=stop_on_token_callback,
)
word_session.execute(insert(Themes).values(themes=response))
word_session.commit()
def stop_on_token_callback(token_id, token_string):
"""
Function to limit return length of the
gpt4all response. Period indicates a sentence.
"""
if "." in token_string:
return False
return True
# def every(__seconds: float, func, *args, **kwargs):
# while True:
# func(*args, **kwargs)
# await asyncio.sleep(__seconds)
#
#
# def main():
# asyncio.ensure_future(grab_word())
# with app.app_context():
# loop = asyncio.get_event_loop()
# loop.run_until_complete(grab_word())
# loop.run_forever()
# async def get_theme(random_word):
# print("Get Theme function running.")
# # SYSTEM_TEMPLATE = "A single sentence based on a word."
# # PROMPT_TEMPLATE = "### Instruction: {0} \n### Response: "
# random_word = await random_word
# print(f"Theme Func, word: {random_word}")
# while True:
# response = MODEL.generate(
# f"Tell me about {random_word}.", temp=0.7, callback=stop_on_token_callback
# )
# with Session(engine) as thm_session:
# print(f"Theme func, response: {response}")
# thm_session.add(response)
# thm_session.commit()